«Behavior Genetics, Vol. 14, No. 2, 1984 Marital Assortment for Personality Dispositions: Assessment with Three Different Data Sources David M. Buss I ...»
Behavior Genetics, Vol. 14, No. 2, 1984
Marital Assortment for Personality Dispositions:
Assessment with Three Different Data Sources
David M. Buss I
Received 15 Mar. 1983--Final 23 Sept. 1983
This study examined spouse correlations in a 9 of 93 married couples
with respect to 16 interpersonal dimensions using three different data
sources: self-report, spouse ratings, and independent interviewer-observer ratings. Results across all three sources supported the previously obtained low positive correlations between spouses. Partial correlations using age and hierarchical multiple regressions using length of marriage do not support the alternative hypothesis that obtained spouse correlations are due to age, to cohort, or to convergence over the course of marriage. Initial assortment is implicated as a probable cause of obtained spouse correspondence.
KEY WORDS: assortative mating; Personality ratings.
The author thanks Michael Barnes, Patricia French, Mary Gomes, Steve Kelner, Ping Lee, and Helen Van Metre for help with conducting this study. Additional thanks go to Michael Barnes for lending his expertise at all stages of this project.
i Department of Psychology and Social Relations, Harvard University, 33 Kirkland Street, Cambridge, Massachusetts 02138.
0001-8244/84/0300-0111503.50/0 9 1984 Plenum Publishing Corporation 112 Buss The study of marital assortment is also important for personality and social psychology, and the consequences of this process are receiving increasing attention (Buss, 1983, 1984; Cattell, 1982; Cattell and Nesselroade, 1967; Eaves and Heath, 1981; Nance et al., 1981). Perhaps one of the most striking ways in which adults create extensive and enduring environments which they subsequently occupy is through the selection of a mate. If that selection is nonrandom, based on personal characteristics, it becomes an important mechanism by which correlations between persons (or genotypes for heritable traits) and environments are created (Buss, 1984; see Plomin et al., 1977b, for a detailed discussion of genotype-environment correlation). Thus, in developing an interactional psychology, the Study of spouse Similarity can illuminate the domains within which person-environment correlations occur, as well as the mechanisms (e.g., initial assortment) responsible for creating such correspondences.
Although systematic meta-analyses have not yet been conducted on marital assortment for personality characteristics, several studies have reported low but consistently positive correlations for a variety of dispositions using self-report personality scales (e.g., Ahem et al., 1982;
Jensen, 1978; Price and Vandenberg, 1980; Vandenberg, 1972). Because validity coefficients for self-report personality scales are always below unity, often dramatically so, convergent evidence of marital assortment from alternative data sources would lend credibility to existing findings.
The present study was conducted to assess spouse similarity in personality characteristics, with particular reference to interpersonal dispositions, using three different data sources to assess dispositions: self report personality measures, spouse ratings, and independent interviewer-observer ratings. The self-report measures consist of some that have previously been employed [e.g., Eysenck Personality Questionnaire (Eysenck and Eysenck, 1975)] as welt as the battery of scales for which no data on spouse similarity have yet been reported [e.g., California Psychological Inventory (Gough, 1957/1964); Interpersonal Adjective Scales (Wiggins, 1979); Personal Attributes Questionnaire (Spence et al., 1979)]. The spouse ratings consist of Likert-type endorsements using seven-place scales on the 16 interpersonal dimensions assessed by the Wiggins (1979) IAS scales. The observer ratings, using composites based on two interviewers' ratings, also entail judgements of the above 16 interpersonal dimensions.
In sum, assessments of the same 16 interpersonal dimensions from three different data sources permits a direct test of whether convergent evidence of spouse similarity can be found with non-self-report measures of personality. Various cognitive, physical, and background variables Marital Assortment 113 were included to establish comparability between the present study and those previously reported.
METHOD Subjects One hundred eighty-six (186) individuals composing 93 married couples participated in the study. Subjects were obtained by placing advertisements and flyers throughout the larger Boston area. Both indicated that a study was being conducted using married couples and that personal feedback and a small sum of money would be given as tokens of appreciation for participation. These and similar methods of recruitment used in studies of married couples may create problems of bias or unrepresentativeness. For example, it is possible that couples who enroll in such studies may be more similar than are couples who do not enroll or that there are biases in attrition over time such that divorces are more prevalent among couples who are least similar (cf. Cattell and Nesselroade, 1967).
Materials Among a larger battery of tests and measures were the following assessment tools used for the present study.
Confidential Biographical Questionnaire. This questionnaire asked a variety of questions about physical characteristics, demographic characteristics, consumption habits, and background marital information. Of particular importance for the present study, in order to establish comparability to published studies on assortative marriage and spouse similarity, were the variables of age, height, weight, handedness, number of siblings, sleep habits, and Consumption habits.
The General Vocabulary Test. To estimate comparability between the present sample and published studies on assortment for cognitive abilities, a multiple-choice vocabulary test (Gough and Sampson, 1974) was completed by subjects.
Self-Report Personality Measures. A battery of personality tests was completed: the Wiggins (1979) Interpersonal Adjective Scales (IAS), the California Psychological Inventory (CPI; Gough, 1957/1964), the Eysenck Personality Questionnaire (EPQ; Eysenck and Eysenck, 1975), a subset of scales from the Personality Research Form (PFR; Jackson, 1967), the Interpersonal Dependency Scales (Hirschfeld et al., 1977), the Personal Attributes Questionnaire (Spence et al., 1979), the Machiavellianism scale (Christie and Gels, 1970), the Self-Consciousness scales 114 Buss (Fenigstein et al., 1975, and the California Self-Evaluation Scale--a measure of general self-esteem (Phinney and Gough, 1982).
Spouse Ratings. Each participant rated his or her spouse on the 16 interpersonal dimensions derived from Wiggins' (1979) taxonomy of interpersonal dispositions, using seven-place rating scales.
Interviewer Ratings. Each couple was interviewed by a pair of interviewers drawn from a seven-member team. The interview consisted of a series of questions centered around each participant's reaction to each of the procedures in the battery, whether they felt that the measures accurately assessed their personalities, their suggestions for improving procedures, and their willingness to participate in a follow-up study. Each interview lasted about 30 min. Directly following each interview, the two interviewers independently rated each participant on the 16 interpersonal dimensions cited above using forms structurally identical to those used for the spouse ratings.
Procedure Data gathering occurred in two sessions, separated by several days.
Evening and weekend sessions were arranged to permit flexible scheduling. Each session lasted about 3 h. In the first session, participants completed the confidential biographical questionnaire, the battery of selfreport measures, and other instruments. In the second session, participants completed the vocabulary test, the spouse ratings, and other measures. Interviews took place during the second testing session. Subjects were tested in groups that ranged from 2 (a single couple) to 14 (seven couples). Each couple was separated for the duration of the testing session to prevent discussion of the measures. To facilitate careful completion of the test battery, refreshments were provided and subjects were en, couraged to take breaks to combat boredom and fatigue.
RESULTS Spouse Correlations for Age and Background Variables
To establish comparability between the findings of the present study and those reported in the behavioral genetics literature, Table I shows descriptive statistics and spouse correlations for age, physical variables, handedness, number of siblings, average amount of nightly sleep, smoking, drinking, and tested vocabulary. The spouse correlations for age, height, and weight are comparable to those found in previous studies (see Spuhler, 1968, for a summary of previous studies). Quite high correlations Marital Assortment 115
" Self-reported smoking was recorded on a seven-place scale and coded as follows: none (0), 1-5 cigarettes per day (1), 6-10 per day (2), 11-20 per day (3), 1.5 packs per day (4), 2 packsper day (5), and 3 packs per day (6).
b Drinking was recorded on a six-place checklist and coded as follows: none (0), occasional drink (1), few times per week (2), 1-2 per day (3), 3-4 per day (4), and 5+ per day (5).
~ Self-reported in response to "Are you left- or right-handed?" d Self-report in response to "Are you a 'day' or a 'night' person?" *P 0.05.
** P 0.01.
*** P 0.001.
were computed, controlling for spouse's ages, between spouses for each of the variables shown in Table I (height, weight, handedness, and so on). All age-adjusted correlations were within 0.03 correlation point of the unpartialed correlations, a finding highly similar to other studies in the behavioral genetics literature. This suggests that cohort effects can be ruled out as a hypothesis in accounting for the observed correlations between the spouses on these variables, In sum, the results in Table I suggest a strong comparability between the results typically reported in the assortative marriage literature and the results of the present study.
Spouse Correlations for 16 Interpersonal D i m e n s i o n s Using Three Different Data Sources The primary purpose of this study was to assess the degree of spouse
similarity in interpersonal dispositions using three different data sources:
self-report, spouse ratings, and independent interviewer ratings. Table II shows these results.
To control for potential differences among interviewers in use of the rating scales, s c o r e s were standardized f o r each interviewer separately Marital Assortment 117 before compositing (with unit weighting) the ratings for each interviewer pair. The alpha reliability coefficients for these interviewer composites are shown in the rightmost column in Table II. While not strikingly high, these reliabilities are typical of those found between two raters (Wiggins, 1973), and any unreliability serves to attenuate the magnitude of the obtained spouse correlations.
Since the IAS scales were developed only recently (Wiggins, 1979), no data on spouse similarity have yet been reported. Perusal of Table II reveals that the magnitudes of spouse correlations are generally low but positive, with a few notable exceptions. The mean spouse correlations across all interpersonal dimensions are 0.12 for the self-report IAS scales,
0.15 for the spouse ratings, and 0.14 for the interviewer ratings. Thus, these results provide independent support for the previously obtained findings of low positive spouse correlations in the personality realm, although correlations for behaviorally based or act frequency measures of personality (Buss and Craik, 1983a,b) may be slightly higher (Buss, 1984).
Three additional findings warrant note. First, the age-adjusted (through partialing) spouse correlations show an average difference from the unadjusted correlations of only 0.01 correlation point. Second, t h e reliabilities for the interviewer ratings, :although typical of those that employ two raters, are sufficiently low that correlations would be improved nontrivially by correcting for attenuation. Indeed, the obtained correlations between husbands and wives found here may be generally viewed as lower-bound estimates of spouse similarity, since some degree of unreliability in all measures serves to attenuate obtained correlations.
Third, an intriguing anomaly to the generally positive correlations is found with respect to the dimensions of dominance and submissiveness, which both yield negative spouse correlations for the spouse and interviewer data sources yet positive correlations for self-reports. This anomaly might stem from alternative reference groups used. When asked to rate how dominant one's spouse is, the referent might be relative dominance within the marriage itself. Similarly, the interviewers had as their sole data base the spouses' behavior within the interview context, and hence their ratings also probably reflect relative spouse dominance within that context. In contrast, self-judgements of dominance and submissiveness are probably made with reference to the general population or peer group. The pattern of findings here might simply reflect dominant-submissive complementarity within the context of the relationship itself but similarity on these dimensions with respect to the population generally.
In sum, the spouse similarity results from the three different data sources generally support the low but consistently positive correlations 118 Buss found in previous studies that have employed only self-report measures.